Title: RSA Interop 2005 Guidelines
Hi Tom,
Thanks for the feedback. Please see my comments
below.
Everyone,
I have flagged a few items that I would like additional
feedback on. Please respond or I'll bring them up on the
call.
Dave Silver or Terry McBride,
Would you please address Tom's question
below?
14. Section 4.2 (a question for Enspier). It's not clear
to me how the eAuth site knows which IDP the user has selected. Are they
using persistent cookie? I.e., I'm asking this to understand what can or cannot
be done at the SP. I.e,. if we do the eAuth use case for a couple of SPs and
then do the logout. Then a new user tries the demo, how does the eAuth portal
know this is a new request (i.e., the user is no longer logged on to the
intial IDP)? It would seem that the link back to the eAuth site would require a
CCID=<idp entity id> parameter much like the one that is sent to the
SP. Section 4.2.2 mentions using cookies to avoid propagation (but I don't
think this will work).
Regards,
Bob
From: Thomas Wisniewski
[mailto:]
Sent: Wednesday, January 19,
2005 7:18 AM
To: Ciochon, Robert; samldemotech
Cc: Mark
Joynes
Subject: RE: RSA Interop 2005 Guidelines
Bob,
nice job putting this together. Here are some comments on the use
cases/guidelines.
1. General question for optional use case: Can we allow users to
perform additional federations from an SP. I.e., after the user federates, login
in again, and prior to or after they defederate, they can have the capability to
re-federate with the defederated IDP or they can have the capability to federate
with additional IDPs. Obviously this is something that we would take them thru.
Similar to defederation, which is something that might be done after a user
leaves the demo site, additional federations could be left up to each
vendor.
[bobc] I
think we are all concerned with introducing too much complexity into the demo
scenarios due to limited amount of time we will have with each customer and the
amount of time to test and verify each scenario. The optional use
case was designed to demonstrate federation clearly and simply. Could
someone show more? Sure, the tools are all there. But I
would prefer to keep things as simple as possible at this
stage.
2. Section 2.1 bullet 5. I'm not sure if this bullet is necessary (i.e.,
bullet 6 is enough). I.e., the use case is always from a solicited SP
AuthnRequest and not initiated at the IDP.
[bobc] I'll leave it in.
Better too much...
3. Section 2.2 bullet 7. The text
says "Initiate a redirect to...". I think you mean to say "Provide a link
to..."
[bobc] I will
change that.
4. Section 2.2 bullet 8. We should
discuss, for the federation case, when the user is required to log
into the SP. I.e., this can be before the IDP login or after. Are we saying it's
up to the SP or does it always have to be after the IDP login.
[bobc] We can discuss this
further on the call to see what people agree
on.
5. End of section 2.2. You added a
possible requriement of displaying the xml documents that were exchanged. Do
users really want to see this? Also, what are you thoughts on how this is
presented (e.g., thru an external screen/log, or on the actual user browser
screen)?
[bobc] This
was a carryover from last year and is up to each vendor if or how they display
this. It is basically a trace log to lend some credence to what is
being shown.
6. Section 3.1 first sentence. You
say four, but I think you mean three (i.e., login, logout, and eAuth). I think
that breakdown works the best. I think you should move section 4.2 after the
logout discussion so it will be section 4.2. Then change sections 4.3 and 4.4
into 4.2.1 and 4.2.2. Basically following the same format/structure you defined
in the summary section 3.1.
[bobc] I actually meant SP initiated login, idP initiated
login, SP initiated logout and idP initiated logout as the 4 base
SAML cases, and then eAuth as a separate 5th case. I'll have a look
at the section setups when I update it next
time.
7. Section 3.1.3, item 2b. s/on the
SP is/on the SP and is/
[bobc] WIll change
that.
8. General comment on the optional
use case. I beleive the SP should state which name id format it is requesting,
i.e., x509 subj name or a persistent id. This is done via the NameIDPolicy
element. This implies that the SP site, which lists the IDPs that support
persistent identifiers, needs to be smart enough to distinguish between the 2
types of identifiers when calling and IDP. We had said this would be based on
the user name. However, if the answer to (4) above is that the user logs into
the IDP before the SP, then the SP does not know who the user is and therefore
cannot be sure which NameIDPolicy to use. I'm proposing the SP has additional
smarts in it (each vendor does this however they want) so that the IDP links on
the SP site hare two-fold. There are links for the base case IDPs and links for
the optional IDPs. Additionally, the eAuth use case always implies the base case
IDPs, hence x509 Ssubject names. Section 6, item 1, wording would be
updated accordingly.
[bobc] I was heading in that direction myself on the last
call when people said they could just figure things out from the user
name. Let's discuss this some more on the call today since I think your
method will simplify things for the demo
staff.
9. Section 3.2, item 4 and 5. I
don't believe this requirement was mentioned on any call/email. I.e., obtaining
user consent is up to each individual provider. I suppose these could be
optional steps but they should not be required. I.e., when a user clicks on
the IDP link at the SP, they are consentng to federation. The same goes for
section 6 bullet 4.
[bobc] I think it was mentioned in passing when
discussing this case and was never explicitly called out. I know this was
a big deal in the Liberty Alliance specs. Let's discuss on the
call.
10.
Section 3.2, item 8. s/SP triggering/SP via the IDP triggering/ The idea
being the user does another saml login at the IDP by initiating it at the SP
(basically a repeat of your item 1). To me it seemed like the text in
item 8 is suggesting that perhaps the user logs in at the SP.
[bobc] That is more accurate
and will be
changed.
11. Section 3.2, item 9. s/on the
web page/on the SP web page/
[bobc] I will change
this.
12. Section 4.1.2, item 3. s/of
links to the SP resources/of SP resources/ ^ s//(URLs)/ The text suggest
the links (actual URLs) are to the SP site when they are actually urls back to
the IDP with some TARGET=xxx or RelayState=xxx query parameter.
[bobc] I will change
this
13. Section 4.1.2, item 4.
s/resource links and/resources and/
[bobc] I will change
this.
14. Section 4.2 (a
question for Enspier). It's not clear to me how the eAuth site knows which IDP
the user has selected. Are they using persistent cookie? I.e., I'm asking
this to understand what can or cannot be done at the SP. I.e,. if we do the
eAuth use case for a couple of SPs and then do the logout. Then a new user tries
the demo, how does the eAuth portal know this is a new request (i.e., the
user is no longer logged on to the intial IDP)? It would seem that the link back
to the eAuth site would require a CCID=<idp entity id> parameter much like
the one that is sent to the SP. Section 4.2.2 mentions using cookies to
avoid propagation (but I don't think this will work).
[bobc] Moved up for
reply.
16. Section 5.1 item 1b. The Saml spec requires subject confirmation
data, s/MUST NOT/MUST/ and the data should include recipient,
notOnOrAfter, etc..... In general I think you can remove section 3. This
is just a restatement of what is mandatory in the saml spec. I.e., similar
to saying you need to include an IssuerName, etc....
[bobc] Some carryover from
last year's doc. I'll take it out if people find it
confusing.
17. Along the lines of S3ction 5.2.1 item 1, an AuthnRequest should
either not inlcude an RequestedAuthnContext element or should value this
with run:...:Password.
[bobc] Anyone else have feedback on this one?
It is a carryover as
well.
18. Section 5.2.1, item 2
and section 7.1, item 3. Why are there any restrictions at all on
SubjectLocality? Shouldn't this just be up to the IDP.
[bobc] I gather this
was added last year because people had some problems with DNSName.
Anyone else have feedback on this
one?
19. Section 5.2.2, item
1. Is everyone ok, with the 3 attributes. It seems like using postal
address (someone else suggested this attribute at a prior meeting) would add to
the personalization features of the demo.
[bobc] Anyone else have
feedback on this?
20. Section 8.1.1, bullet 1 and 3 AND section 8.1.2 bullet 2.
I thought we agreed that there would be one SP resource for each SP and each
vendor would only have one IDP. So change the zero or more reference to
there is one of these per vendor. Anything beyond one would complicate the eAuth
site, the links necessary on each SP/IDP, etc...
[bobc] Not every vendor is
required to provide both an SP and an idP. As I recall, someone
mentioned that last year not all vendors provided both. That is why the
zero. I will change the "more" to
"one".
21.
Section 8.1.1, bullet 3. s/URL to the idp logon page/The SAML entity id of
the IDP/ The entity id format is like a url but it is NOT a url and is not
meant to be the idp login url.
[bobc] I was using the recommendation in the spec, especially
since we are only using front channel bindings. I was trying to make
it simple, but can change it if it is not accurate for
our demos.
22. Section 8.1.2
bullet 3. Actually remove the IDP entity id reference from 8.1.1 and move it
here, note that this is a URL but an entity id
only.
23.
Section 8.1.2 bullet 1. For issuer name, is this a string for display
purposes?
[bobc] I
believe that is how it was used last
year.
24.
Section 8.1.4. I would say copy the list of vendors into A.2 as well. I.e.,
store the value for <Vendor> in the appendix so it's easy to retrieve.
Also, in A.2 perhaps store the following data point: IDP-Basic, SP-Basic,
IDP-Optional, SP-Optional to indicate which vendor is doing what. It would be
really helpful to nail this down before the dry run.
[bobc] I plan to add that in
somehow once I have who is supporting
what.
25.
Section 8.1.4, I would suggest sticking with the same password convention as the
base case. I.e., the password is the same as the userid. It just makes things
easier I think.
[bobc] Or should we make just one password that
is the same for all users for both base and optional? Anyone
else have feedback on
this?
26.
Section 9.2.2 bullet 5 s/qill/will/
[bobc] I will fix
that.
27. Section A.2
and 8.1.1. The additional URLs that are necesary (in A.2 and in 8.1.1)
include:
-
SingleSignOnService (at IDP for base and optional
case)
-
AssertionConsumerService (at SP for base and
optional case)
-
SingleLogoutService (request) (at IDP and SP for
base and optional case)
- SingleLogoutService
(response) (at IDP and SP for base and optional
case)
- ManageNameIDService (request)
(at IDP and SP for optional case
only)
-
ManageNameIDService (response) (at IDP and SP
for optional case only)
[bobc] Will expand those
sections
28. Section A.2
and 8.1.2. The additional identifiers that are necessary (in A.2 and in 8.1.2)
include:
-
SP entity id (simliar to idp entity
id).
-
Receiver Name (similar to the idp issuer name). I.e., a display name for the
SP.
[bobc] Will add
receiver name. SP entity id is same question as
idP.
Tom.
-----Original Message-----
From: Ciochon, Robert
[mailto:]
Sent: Tuesday, January 18, 2005 3:17
PM
To: samldemotech
Subject: RSA Interop 2005
Guidelines
Hi,
Attached is
the first cut at defining the guidelines and use cases for the SAML Interop at
RSA 2005. Please send me any feedback, questions, clarifications,
etc...
Regards,
Bob
<<RSA2005-saml-interop.doc>>
Robert Ciochon
eTrust Development Manager
Computer
Associates
San Diego, California
(858) 625-6866
